Abstract: The method for rapid detection of oxytetracycline residues in milk was established by combining mid-infrared （MIR） spectroscopy with partial least squares discriminant analysis（PLS-DA）. Firstly, 62 pure milk samples and 62 milk samples adulterated with different concentrations of oxytetracycline（0.02-30.00 mg/L）were prepared respectively. The MIR spectra of all samples were collected in the range of 400-4 000 cm^-1. After the pretreatment of original spectral data, the PLS-DA models were established in the full band and sub band ranges, respectively. The results show that different wave-number ranges had a great influence on the prediction model of milk adulterated with oxytetracycline. The classification accuracies of the constructed PLS-DA models were 100% for calibration set in the full band and sub band ranges, and the correlation coefficients of models were between 0.995 and 0.997 for prediction set, which shows that the models had a good fitting effect. The classification accuracies of constructed models in the full band and in the range of 2 000-4 000 cm^-1 were 90.0% for prediction set.

Key words: milk, antibiotics, mid-infrared spectroscopy, partial least squares discriminant analysis